Cranky Thoughts On AI...

 The episode in which your regular Cranky Correspondent muses rather crankily on the current state of AI and where it could really make a difference...

I know, I know, but the thing is We Are Not Yet getting Artificial Intelligence.

There are many ways to define "Artificial Intelligence".  The classic one is when you hold a conversation with something which you believe is human, and can't tell the difference.  I'll give you a little more complex one.  True Artificial Intelligence will be able to respond to a situation it has not encountered or prepared for and selects an appropriate response.  

What we have today are expert systems.  The way you know that is because they are trained.  They are exposed to situations others have experienced and the things that others have done to respond to them.  They then develop their response based on previous responses.  So yes, I have also just described much of what some people consider training.  It is not what I would call Education, but it has become what some people and some systems consider adequate.  I do not.

When you have a small child, you teach them using very simple if/then instructions.  If you see something on the stove, stay away, it is hot.  Simple computer programming is a lot of If/Then statements.  But that's a reaction, not a pro-active system.  

 A truly intelligent system could be left alone to deal with things with minimal supervision.  

Like, for example, the number of daily phone calls I get on my home phone.  For some reason, some assholes think they're going to accomplish something by calling and leaving a message for me about the loan I applied for.  Funny thing is that I have not applied for a loan on about 25 years.  And if I ever do, I guarantee you it will be with someone local, not someone with an 800 number or non-local.  And I will find a way to have written into the terms of the loan that the iterest rate will be cut by 50% every time the loan is sold to a new servicer, who will then be required to pay 10% of the current outstanding balance to me for the inconvenience of having to deal with them.

Because, you see, no sane individual will ever agree to any such proposal, so I won't ever get a loan.  

But the reason I know that none of these calls is from a legitimate financial institution is because the calls are almost always automated, there is no human on the other end of the call, and they do not require a call back or follow up.  They are all pretty much the same.  They have found ways to put messages into my voice mail box without calling, the messages almost never start at the tone, and are rather typically in mid-word when the recording starts.  And they are obviously being paid by the number of messages they leave, not verified or confirmed valid customer contacts.

So when people start talking about AI, no, we're not there yet.  You want to impress me?  Show me an AI system that can accept calls, and if it is from an organization that says they're responding to my request for a loan, they need to ask the caller to confirm the date of the signed agreement, and the middle name of their mother.  Since they don't exist, no problem.

Or build a filter for spam.  Really simple - the email comes in from a non-existent server - delete it.  The email comes from an obviously fake email address that is a random collection of letters and/or numbers?  Trash.  What's that, you say you use an email address like that?  Well, that's cute.  I don't want your email either.  I know the internet email protocol requires forwarding of all messages, but I do think that it's beyond time for us to amend that protocol to state that if an address is not from a pre-published "white list" where the user has explicitly permitted the address to email them, the sender must either show an identifer which the receiver has confirmed is the one they send to others to allow them to initiate contact, or they can send a pre-set amount of money ito an account that they will forfeit if they were not permitted.  Seems fair.